I don't really understand the benefit reprojection provides. Why not just show the user the same frame if the next one isn't ready to be rendered yet?
anavmehta12
reprojection is useful cause it reduces the delay between when you move your head and when you actually see the image. If we move our head slightly to the left or right instead of rendering the scene again it just reprojects the previous scene to fit our current view.
spegeerino
In this case, is a "frame" rendered larger than the FOV (or even just full 360) so that even after the head movement and reprojection starts the appropriate data is available to make a frame that carries all the info that it should? For example, if I was able to completely turn around between the rendering of two high-quality frames, would the reprojection be able to show me what's behind me before the next frame is rendered or would I be stuck looking at some extremely warped version of what was in front me before?
henrykhaung
This is such a genius solution. If the VR headset cannot generate new frames fast enough, instead of rendering them, we simply take the last frame and adjust it a bit to match your current head position
dhruvchowdhary
Reprojection sounds smart for quick head turns. But, can it handle big moves or only small adjustments before the picture looks weird?
I don't really understand the benefit reprojection provides. Why not just show the user the same frame if the next one isn't ready to be rendered yet?
reprojection is useful cause it reduces the delay between when you move your head and when you actually see the image. If we move our head slightly to the left or right instead of rendering the scene again it just reprojects the previous scene to fit our current view.
In this case, is a "frame" rendered larger than the FOV (or even just full 360) so that even after the head movement and reprojection starts the appropriate data is available to make a frame that carries all the info that it should? For example, if I was able to completely turn around between the rendering of two high-quality frames, would the reprojection be able to show me what's behind me before the next frame is rendered or would I be stuck looking at some extremely warped version of what was in front me before?
This is such a genius solution. If the VR headset cannot generate new frames fast enough, instead of rendering them, we simply take the last frame and adjust it a bit to match your current head position
Reprojection sounds smart for quick head turns. But, can it handle big moves or only small adjustments before the picture looks weird?